Machine Learning Classification of One-Chiral-Center Organic Molecules According to Optical Rotation

Abstract

In this study, machine learning algorithms were investigated for the classification of organic molecules with one carbon chiral center according to the sign of optical rotation. Diverse heterogeneous data sets comprising up to 13,080 compounds and their corresponding optical rotation were retrieved from Reaxys and processed independently for three solvents: dichloromethane, chloroform, and methanol. The molecular structures were represented by chiral descriptors based on the physicochemical and topological properties of ligands attached to the chiral center. The sign of optical rotation was predicted by random forests (RF) and artificial neural networks for independent test sets with an accuracy of up to 75% for dichloromethane, 82% for chloroform, and 82% for methanol. RF probabilities and the availability of structures in the training set with the same spheres of atom types around the chiral center defined applicability domains in which the accuracy is higher.

Impact of conformation and intramolecular interactions on vibrational circular dichroism spectra identified with machine learning

“…We want to extract a VCD spectrum solely from a conformer geometry. Figure 1 contrasts the current work against our previous work 52 and other recent works 53 – 55 that address the link between an AC and an experimental spectrum or property. The present paper concentrates on the link between the structure of a conformer and its VCD spectrum within a given AC of a molecule.…”

Learning 3D Representations of Molecular Chirality with Invariance to Bond Rotations

“…Explicit representations of chirality in machine learning models. A number of machine learning studies account for chirality through hand-crafted molecular descriptors (Schneider et al, 2018;Golbraikh et al, 2001;Kovatcheva et al, 2007;Valdés-Martiní et al, 2017;Mamede et al, 2021). A naïve but common method for making 2D GNNs sensitive to chirality is through the inclusion of chiral tags as node features.…”

“…Following Pattanaik et al (2020a), we also employ a toy R/S chiral label classification task as a necessary but not sufficient test of chiral recognition. For a harder classification task, we follow Mamede et al (2021) in predicting how enantiomers experimentally rotate circularly polarized light. Lastly, we create a dataset of simulated docking scores to rank small enantiomeric ligands by their binding affinities in an enantiosensitive protein environment.…”

“…Optical activity is an experimental property, and has no correlation with R/S classifications (Table15, appendix A.7.2). FollowingMamede et al (2021), who used hand-crafted descriptors to predict l / d labels, we extract 30K enantiomers (15K pairs) with their optical activity in a chloroform solvent from the commercial Reaxys database(Reaxys, Accessed Aug. 6-9, 2021), and generate 5 conformers per enantiomer using the ETKDG algorithm in RDKit…”